Reproduce “The CARFAC v2 Cochlear Model in Matlab, NumPy, and JAX” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Richard F. Lyon, Rob Schonberger, Malcolm Slaney, Mihajlo Velimirović, Honglin Yu

Google researchers updated the CARFAC cochlear model to v2, addressing previous physiological anomalies by relocating the AC-coupling filter and refining the Inner Hair Cell model. The updated model is now available in MATLAB, NumPy, and JAX, with the JAX implementation offering significant speed improvements and automatic differentiation capabilities for integration into machine learning workflows.

Key insights

  • Refining the AC-coupling filter's placement effectively removes non-physiological DC anomalies from Basilar Membrane output while preserving desirable low-frequency distortions.
  • A new two-capacitor Inner Hair Cell (IHC) model significantly improves physiological realism by reducing neural synchrony at high frequencies.
  • Implementing CARFAC in JAX provides substantial computational speedups and enables automatic differentiation, allowing for gradient-based optimization and seamless integration into machine learning pipelines.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.